[SOLVED] Cannot seem to NFS mount Btrfs subvolumes

rnturn · 04-02-2018, 10:04 PM

Background:

Server: openSUSE Tumbleweed (20180326. Yep... newly updated)
Client: openSUSE 13.2

I have a couple of btrfs subvolumes mounted on the server system similar to:

Code:

/data/test          (mount point in ext4 filesystem)
        +---test1   (mount pt. for btrfs subvolume)
        +---test2   (mount pt. for btrfs subvolume)

I tried to export these two (and more in the future which is why I'm wanting to have a single record in /etc/exports) via NFS by exporting them using:

Code:

/data/test clientsys(ro,nohide,no_subtree_check)

What I'm seeing when I mount on the client using:

Code:

mount -t nfs -o ro,nfsvers=4,tcp,rsize=8192,wsize=8192,retry=5 server:/data/test /opt/app/data

is a message

Code:

Mounting NFS filesystem: l:l

while the CPU utilization immediately goes to 100% (good thing there's two), and the mount command fails to return to a shell prompt.

I can go to another window and see that the client has mounted the remote filesystem but I can't find anything under the two mount points. If I issue "^C" on the mount command, it's still mounted but, again, no data is visible below the mount points. It appears that the "nohide" option doesn't do what I thought the manpage said it was supposed to do. Or... the client's NFS implementation (nfs-client-1.3.0-4.4.1.x86_64) may be too old---there is a warning about that possibility on the manpage.

Does anyone know if my client system OS is just incapable of dealing with mounting Btrfs via NFS? I can pretty easily revert back to using one huge filesystem with "test1", "test2", etc. being subdirectories for the time being--in fact, I'm doing it now--but it'd be nice to take advantage of btrfs's features.

Thoughts? Ideas? Pointers? All are welcomed.

--
Rick

Additional info: My Raspberry Pi was able to mount the Tumbleweed NFS export just fine though this may be due to it having had a complete set of updates applied 2-3 weeks ago.

voleg · 04-04-2018, 04:36 AM

Comments:
1. When exported via NFS, no child mounts exported recursively. You have to export them explicitly.
2. When exporting via NFSv4, you have to use fsid=XX option with different fsid numbers for different exports.
3. BTRFS subvolumes are different inode space, they have to be exported separately.

Probably you have other problem, but these are must to fix.

rnturn · 04-04-2018, 04:58 AM

Quote:

Originally Posted by voleg

Comments:
1. When exported via NFS, no child mounts exported recursively. You have to export them explicitly.
2. When exporting via NFSv4, you have to use fsid=XX option with different fsid numbers for different exports.
3. BTRFS subvolumes are different inode space, they have to be exported separately.

Probably you have other problem, but these are must to fix.

Thanks. I was so hoping my interpretation of the description of the "nohide" option would eliminate the need to explicitly define a record in /etc/exports for each subvolume I want to mount---as well as an addition record in the client's /etc/fstab. But... as you reminded me, the inode spaces would be a problem.

Is there an option for marking this thread [Unsolvable]? :/

Thanks again...

voleg · 04-05-2018, 03:11 AM

Nohide option should be part of child mount exportfs, not parent as you did.
In addition, even "man" page of this option mentions, that NFS client can be confused detecting duplicate inodes in "same" filesystem.

rnturn · 04-06-2018, 05:04 AM

Quote:

Originally Posted by voleg

Nohide option should be part of child mount exportfs, not parent as you did.
In addition, even "man" page of this option mentions, that NFS client can be confused detecting duplicate inodes in "same" filesystem.

The mounts are failing to complete even after reverting to a single exported filesystem (as opposed to the "tree" of subvolumes I was hoping to use). It seems there are plenty of new wrinkles to work out when mixing NFS3 and NFS4. Always something new to learn. I'm putting this problem on the back burner for now anyway; too many other things to fix/clean up on this rebuilt server.

Cheers...